[ntp:questions] Re: Dopey selection of server
David L. Mills
mills at udel.edu
Sun Feb 1 04:57:44 UTC 2004
Don't trust the tinker floor command until you make sure it does what
you want; it was designed for the manycast scheme, not for the purpose
Assuming the network path you have has no dial-up or ISDN circuits, the
delay you find on the first message is probably an ARP thing. Proteon
routers used to overwrite the packet buffer with an ARP if the cache was
bare. In other words, if the poll interval ramped over the ten-minute
ARP cache timeout, no packets got through at all. Hmm. That was fifteen
years ago; now Proteon is gone and hopefully so is that ugly ARP
The burst and iburst subcommands do have provision to vary the interval
between the first and second packets for the very purpose you note.
Christopher Burian wrote:
> Dave, thanks for the in-depth reply. I do appreciate it and I know you
> have better things to do.
> It sounds like tinker floor is just what I'm looking for, a way so that
> my client doesn't pester stratum 1 servers unless it's stratum 2 servers
> are OTL.
> Changing the subject... I wonder if this would actually do any good or
> not as far as jitter and delay... but using a Windows SNTP client, I've
> noticed that querying a server results in a long delay on the first try,
> but is almost instantaneous on the second try. It would be nice to be
> able to double-tap each server and then discard the first reply at each
> poll interval. Kind of like burst.
More information about the questions